전자신문 기사 일부 인용입니다.
https://www.etnews.com/20260701000294
앤트로픽이 최신 인공지능(AI) 모델 '페이블5' 서비스를 1일(현지시간) 재개한다.
앤트로픽은 미국 정부와 페이블5에 대한 접근을 복원하기로 합의했다고 30일(현지시간) 밝혔다. 아마존 연구원들이 제기한 사이버 공격이 가능한 '클로드 미토스' 모델의 공개 버전인 페이블5 보안조치 편법 우회 등 문제를 해결하기로 하며 수출 금지 해제 실마리가 풀렸다. ……
한편 앤트로픽은 이날 신규 에이전틱 AI 모델 '클로드 소넷5'를 출시했다. 스스로 계획을 세우고 브라우저·터미널 등 도구를 활용해 자율적으로 작업을 수행하는 능력을 향상시켰다.
추론, 도구 사용, 코딩, 지식 업무 등 에이전트 성능 핵심 영역에서 전작 '클로드 소넷 4.6' 대비 성능이 개선됐고 상위 모델 '클로드 오퍼스 4.8'에 근접한 성능을 보다 낮은 가격으로 제공하는 게 특징이다. [나머지 내용 생략]
https://x.com/AnthropicAI/status/2072163884430229756
Ini adalah kutipan dari artikel berita Elektronik.
https://www.etnews.com/20260701000294
Anthropic akan melanjutkan layanan model kecerdasan buatan (AI) terbaru 'Claude 5' pada tanggal 1 (waktu setempat).
Anthropic mengumumkan pada tanggal 30 (waktu setempat) bahwa mereka telah mencapai kesepakatan dengan pemerintah Amerika Serikat untuk memulihkan akses ke Claude 5. Masalah yang ditimbulkan oleh peneliti Amazon, seperti pelanggaran keamanan dan metode penyimpangan dalam versi publik model 'Claude Mithos' yang rentan terhadap serangan siber, akan diselesaikan. Ini membuka jalan bagi penghapusan larangan ekspor.
Sementara itu, Anthropic meluncurkan model AI agen baru 'Claude sonnet 5' pada hari ini. Kemampuannya untuk merencanakan sendiri dan menggunakan alat seperti browser dan terminal untuk melakukan tugas secara mandiri telah ditingkatkan.
Kinerja agen di bidang inti seperti penalaran, penggunaan alat, pengkodean, dan pekerjaan berbasis pengetahuan telah ditingkatkan dibandingkan dengan model sebelumnya 'Claude sonnet 4.6'. Fitur utamanya adalah menawarkan kinerja yang mendekati model tingkat atas 'Claude Opus 4.8' dengan harga yang lebih rendah.
https://x.com/AnthropicAI/status/2072163884430229756
This is a partial quote from an article in the Electronic Newspaper.
https://www.etnews.com/20260701000294
Anthropic is resuming service of its latest artificial intelligence (AI) model, 'Claude 5,' on July 1st (local time).
Anthropic announced on June 30th (local time) that it had reached an agreement with the US government to restore access to Claude 5. The company agreed to address issues raised by Amazon researchers, such as bypassing security measures in 'Claude Mithos,' a publicly available version of Claude 5 that could be vulnerable to cyberattacks. This paved the way for lifting the export ban.
Meanwhile, Anthropic launched a new agent AI model called 'Claude sonnet 5' on the same day. It has improved its ability to plan independently and use tools such as browsers and terminals to perform tasks autonomously.
Performance in key areas of agent performance, such as reasoning, tool use, coding, and knowledge work, has been improved compared to the previous model 'Claude sonnet 4.6.' It offers performance close to that of the higher-end model 'Claude Opus 4.8' at a lower price.
[The rest of the content is omitted]
https://x.com/AnthropicAI/status/2072163884430229756
